# lilt-camembertv2-base

This repository contains **lilt-camembertv2-base** weights exported to `safetensors` format.

## Source

- **Layout encoder**: [`jpwang/lilt-only-base`](https://github.com/jpwang/lilt) — layout-only pretrained checkpoint
- **Text encoder**: [`almanach/camembertv2-base`](https://huggingface.co/almanach/camembertv2-base) — French language model

## Methodology

Produced by **weight merging** following the official [`gen_weight_roberta_like.py`](https://github.com/jpwang/lilt) procedure. No training or fine-tuning was performed.

Intended as a starting point for downstream fine-tuning on French document understanding tasks (NER, token classification, extractive QA…).

## Files

| File | Description |
|------|-------------|
| `model.safetensors` | Model weights |
| `pytorch_model.bin` | Model weights (PyTorch format) |
| `config.json` | Model configuration |
| `tokenizer_config.json` | Tokenizer configuration |
| `tokenizer.json` | Tokenizer (fast) |

## License

Weights are derived from models released under MIT and Apache-2.0 licenses.

## Acknowledgements

- [LiLT: A Simple yet Effective Language-Independent Layout Transformer](https://arxiv.org/abs/2202.13669) — Wang et al., 2022

> **Note**: This is not an official release from any of the above organizations.